Radioactive source calibration test of the CMS Hadron Endcap Calorimeter test wedge with Phase I upgrade electronics

The Phase I upgrade of the CMS Hadron Endcap Calorimeters consists of new photodetectors (Silicon Photomultipliers in place of Hybrid Photo-Diodes) and front-end electronics. The upgrade will eliminate the noise and the calibration drift of the Hybrid Photo-Diodes and enable the mitigation of the radiation damage of the scintillators and the wavelength shifting fibers with a larger spectral acceptance of the Silicon Photomultipliers. The upgrade also includes increased longitudinal segmentation of the calorimeter readout, which allows pile-up mitigation and recalibration due to depth-dependent radiation damage. As a realistic operational test, the responses of the Hadron Endcap Calorimeter wedges were calibrated with a 60Co radioactive source with upgrade electronics. The test successfully established the procedure for future source calibrations of the Hadron Endcap Calorimeters. Here we describe the instrumentation details and the operational experiences related to the sourcing test.

. A picture of the HE (left) and a sketch of the longitudinal and transverse segmentation (right, not to scale). Within the same η tower, longitudinal sections with identical colors (red, green or blue) belong to the same calorimeter readout channel. The dots denote the light injection points for the laser calibration system.
The Phase I upgrade of HE consists of new photodetectors (Silicon photomultipliers -SiPMs) and upgraded front-end electronics (QIE11 -Charge Integration and Encoding [3]). The upgrade will allow the elimination of the high amplitude noise and drifting response of the HPDs, at the same time enabling the mitigation of the radiation damage of the scintillators and the wavelength shifting fibers with the larger spectral acceptance of the SiPMs. The upgrade will also have increased longitudinal segmentation of the readout, which is beneficial for pile-up mitigation and recalibration to correct for depth-dependent radiation damage [4].
As a realistic operational test, the responses of the Hadron Endcap Calorimeter test wedges, permanently located in the CERN H2 beam line in the SPS North area [5], were calibrated with a 60 Co radioactive source with upgrade electronics. Here we describe the nature of the test, discuss the results and operational experience.

Description of the test setup and the data acquisition
The source calibration test was performed with two of the φ sectors of one of the permanently installed wedges in the CERN H2 test beam area. Figure 2 shows a picture of the test area. Test -2 -

JINST 12 P12034
segments of the CMS hadron endcap, barrel and outer calorimeters are installed on a motion table which allows precision alignment of the detectors with respect to the test beam. During the source calibration test, there was no beam in the test area.  Figure 3 shows a sketch of the HE test megatile. The megatile consists of 16 scintillator tiles, all of which have sigma-shaped grooves machined on the surface near the periphery of the tiles. The WLS fibers inserted in these grooves are then individually coupled to clear fibers which transport the light to the readout modules housing the 96 SiPMs. Only two of the four available slots of readout modules in the HE RBX were used in this test. The RBX also houses the control and QIE11-based front-end electronics which are connected to the back-end electronics in the control room via control and data fibers. The back-end electronics is the current µTCA-based system of the CMS HE [6]. Figure 3 also shows the locations of the source tubes inside the megatile as red tubes. The source tube along φ = 5 sector goes across six scintillator tiles, and the one along φ=6 sector goes across ten tiles. Also shown are the η indices from 16 to 25. Although structurally and functionally identical, the geometry of the test megatile differs from the actual megatiles of the HE. In the actual HE megatiles, η = 21 tile has a single φ sector and the η segmentation goes up to 29 instead of 25. Because the test wedges have been connected and disconnected as part of ten years of testing, the performance of the optical connections is expected to have degraded to some extent. Therefore, it was not intended to obtain an absolute calibration as a result of this test. Up to 13 megatile layers were tested with the radioactive source. Various consecutive layers in the wedge were combined to form calorimeter towers. The towers with η indices between 16 and 21 cover 5 • in φ and those with η indices between 22 and 25 cover 10 • angles. Figure 4 shows a picture of the radioactive source driver, the indexer and the source tubes. The radioactive source is loaded at the end of the source wire that is wound around the source driver reel. The source wire is made of stainless steel and has an inner and outer diameter of 0.406 mm and 0.711 mm respectively. The radioactive source, in the shape of a cylinder with 1 mm diameter and 100 µm thickness, is inserted into the end of the wire and held in place by a fine steel piano wire. The outer source wire and inner piano wire are then crimped together at the back end in order to fix the position of the radioactive source. The activity of the 60 Co source used in this test was  A polycarbonate reel, belt driven by a DC reversible electric motor, is used to insert or retract the source wire into or out of the calorimeter source tubes. The calorimeter source tubes are coupled to acetal plastic tubing to mediate the transfer of the source wire from the source driver into the calorimeter. The transition between the plastic tube and the stainless steel source tube inside the calorimeter typically involves small-angle conical holes in brass that channel the source wire to the calorimeter source tube.
The driver system also contains an additional electric motor that functions to select the source tube into which to direct the source wire, an action referred to as indexing. The position of the radioactive source, relative to the source driver and referred to as the reel position, is provided by an optical rotary encoder read out by industrial batch counters. The speeds at which the radioactive source may be inserted or retracted by the driver are between 1 and 12 cm/s. The data acquisition system continuously records the location of the radioactive source and the data from the 96 SiPMs. The data from the SiPMs are histograms of ADC counts for a predefined number of time samples (TS) of 25 ns length each (40 MHz, which is the LHC frequency). The system can also take pedestal and LED data, where the data from the SiPMs are randomly read out for the former, and the data taking is initiated by the firing of LED light into the SiPMs for the latter. In both -4 -

JINST 12 P12034
of these cases, the data is ten consecutive time samples. Due to the ongoing firmware development of the front-end electronics at the time of this test, the data from the first time sample was not available.

Calibration of the SiPMs
For the calibration of the SiPMs, dedicated pedestal data is used. The SiPMs had 15 µm pitch pixels in a 2.8 mm diameter effective photosensitive area, and were made by Hamamatsu [7]. They were operated at a stable temperature of 18 • C provided by peltier cooling. The dark current rate for these SiPMs were approximately 1 MHz. SiPM bias voltages were turned on and were also stable during the pedestal data taking. Therefore, a sufficiently large integration time window in the pedestal data would contain up to a few photon pulses. One can then extract the charge per detected photon (defined as the gain of the SiPM hereafter) from the distribution of the integrated charges.
For each time sample, the raw ADC values read out from the QIE11 front-end cards are converted into corresponding charge values using the conversion tables specific to the QIE11 chips [3]. Of the nine available consecutive time samples, four are used to form the 100-ns integration window. A selection on the sum of the charges in the neighboring time samples is imposed in order to prevent transients and have the entire single or multiple photon pulses contained within the integration window. The integration and the transients windows are shown in figure 5 (left) as red and grey shaded areas.  . Also shown is the Gaussian fit which was used to determine the maximum signal allowed in the transients window (right).
-5 - Figure 6 shows the integration window charge spectrum of this SiPM, using the same dedicated pedestal data. Pedestal and the one, two and three photon peaks are clearly measurable. Four, five and six photon signatures are also observable. The spectrum is fit with a single function: where µ i = µ 0 + iG. The function is basically a sum of Gaussians with independent weights A i and widths σ i ; and means µ i are parametrized as a function of the pedestal mean µ 0 , the peak id i and the SiPM gain G (charge generated by a single photon). The range of the fit is such that all the photon peaks are larger than 1/1000 of the pedestal peak. Figure 6 shows all the parameters obtained for this particular SiPM. The gain of the SiPM is a parameter of this fit function.
Entries 19993 6.9 ± 84.1 0.8 ± 6.8 Figure 6. Example of integration window charge spectrum of the SiPMs. Also shown is the combined fit to the spectrum.
During the radioactive source test, the SiPM bias voltages were adjusted for two gain values of 50 fC and 40 fC. Figure 7 shows the measured gain values for the 50 fC setting (blue) and 40 fC setting (red) for all SiPMs. Also shown are the constant fits to the gain values yielding 50.40 ± 0.01 fC and 40.71 ± 0.01 fC respectively. The uniformity of the gains is sufficient for the purposes of the test. These gain values are used to convert the charge values calculated in the sourcing test into number of photoelectrons as described in the next section.

Analysis of the radioactive source calibration data
The radioactive source calibration data is a set of histograms periodically accumulated by the back-end electronics for each readout channel in ADC units. Also recorded is the position of the radioactive source inside the megatile. Individual histograms are first converted from ADC units to charge in fC. After the pedestal subtraction, the average charge values of these histograms are converted to number of photoelectrons using the previously obtained gain values. It should be noted here that this is a numerical conversion to equalize the reponses of the SiPMs and not an actual calculation of the number of photoelectrons. Each bin of the histogram is an integral for 25 ns (instead of 100 ns in the gain calculation) and the signal pulse is not entirely contained in this 25-ns window. Figure 8 shows the response profiles for φ = 5 (left) and φ = 6 (right) sectors for layers 9 and 8 respectively. The boundaries of the tiles are clearly distinguishable. The profiles are precise enough to calculate the longitudinal size of the tiles as seen by the radioactive source.  In order to obtain an average response value to the radioactive source, the spectrum of the number of photoelectrons is extracted for each signal channel with a zero suppression at 0.07 photoelectrons. This zero suppression value is selected to be slightly above the usual pedestal fluctuations measured of around 0.05 photoelectrons (2.6 fC or σ 0 in figure 6). Figure 9 shows this spectrum for the signal channel of the φ = 5, η = 20 layer 9 tile. The spectrum is fit to a Gaussian -7 -function with a variable width defined as where N 0 is the weight, µ is the mean, σ 0 is the central width, α 1 , β 1 and α 2 , β 2 are the parameters defining the varying width to the left and right of the mean respectively [8]. As long as sufficient statistics is present in the distributions, the fits are successful in describing the spectra. The statistics is determined mainly by the speed of the radioactive source inside the megatile, which can be adjusted as a simple parameter in the data acquisition and control system, and will be discussed further below.
= 20, layer 9 η = 5, φ  Figure 9. Photoelectron spectrum for the signal channel of φ = 5, η = 20 layer 9 tile. The spectrum is fit to a Gaussian function with a variable width defined in the text.

Results
Following the basic electronics-detector mapping test and calibration, various measurements were performed. The efficiency of radioactive source tests in identifying possible issues in the optical paths as well as establishing the scintillating tile aging and radiation damage was studied by comparing the signals from different layers that form the same physical detector tower and are read out by the same SiPM. Figure 10 shows the responses measured in layers 1 and 2 (left), and layers 7 and 8 (right) when the radioactive source was moving in φ = 5 sector. The error bars are smaller than the size of the marker points. Standard errors are calculated for the averages. Layers 7 and 8 show minimal deviation in response to the radioactive source. The overall decreasing trend towards higher η values is due to the decreasing lateral sizes of the tiles. No attempt was made to correct for this effect in this study. The responses in layers 1 and 2 on the other hand show significant (more than 15 %) deviation in η = 18 tower. The major reason is that these layers are the most upstream -8 -

JINST 12 P12034
layers in the stack and have seen several physical interventions over the course of several years. The sourcing test can successfully identify such potential problems at the per tile level.  Figure 11 shows a comparison of the responses for layers 7 and 8 (left), and layers 10 and 12 (right) for the φ = 6 sector. All tiles have responses around 0.5 photoelectrons and the variation of the response as a function of the tile size is within expectations. The deviations of the tile signals between layers 7 and 8 and between layers 10 and 12 are within 15 %. The quality of the results obtained indicates that with periodic radioactive source calibration of the actual HE detector during the Phase I era, the combined performance degradation of the tiles and the optical path due to radiation can be precisely measured and the response can be corrected.
The φ = 5 sector was also subjected to further testing of the radioactive source calibration. The first of these tests was the dependence of the source response on the source speed inside the megatile. As the source moves faster, the number of histograms and hence the number of data points in the photoelectron spectra (e.g. figure 9) are reduced. Therefore, the fit procedure may become problematic. The default source wire speed in these tests was 2 cm/s. φ = 5 sector was also tested at 1 cm/s to search for any improvements in the results. Figure 12 (left) shows the response at 2 cm/s (black) and 1 cm/s (red) source wire speeds as well as the deviations per tile for the tiles in layer 6 in φ = 5 sector. The deviation between these two speed settings is minimal. The overall statistics and the quality of the fits of the photoelectron spectra improve as expected. Since larger speeds are desired for practical purposes, it can be concluded that the final results are not affected by increasing the speed of the source wire by a factor of two, from 1 cm/s to 2 cm/s. For the default setting of 2 cm/s, the radioactive source calibration of one wedge takes around 40 minutes. With the information obtained in these tests, the time and effort needed to set up the calibration hardware and to perform the operations for HE were estimated with sufficient accuracy.
-9 -   Figure 12. Source response of the φ = 5 sector tiles at layer 6 for the source wire speeds of 2 cm/s (black) and 1 cm/s (red) (left); and for the SiPM gain settings of 50 fC (red) and 40 fC (green) (right) together with their per tile deviations (lower panels).
-10 - Figure 12 (right) shows the response of the φ = 5 sector tiles with different bias voltage settings of the SiPMs. The red squares are for the default setting of 50 fC gain, and the green triangles are for the gain setting of 40 fC. Also shown is the deviation per tile in the lower panel. As expected, the response at 40 fC gain is on average 20 % lower than the reponse at 50 fC gain. It can be concluded that the radioactive source calibration results can be translated to a different operating gain of the SiPMs within 5 % accuracy. This information would be useful in the long term in identifying the systematics of the radioactive source calibration due to e.g. the radiation damage to the SiPMs.

Conclusions
The test described here validates the radioactive source calibration operations of the actual CMS Hadron Endcap calorimeter wedges several months ahead of the actual Hadron Endcap Calorimeters source calibration campaign during the Extended Year End Technical Stop of 2016-2017. The test allows the estimation of the time and effort needed for the actual source calibration campaign. It provides insight about the effect of the source wire speed and the SiPM calibration on the quality and quantity of the data.
Various techniques were developed for the analysis of the data. The analysis methods developed during the test combine the features of fast calibration of large-scale SiPM arrays and fast extraction of the radioactive source response of several thousands of scintillator tiles, and were successfully implemented at the actual sourcing campaign. This test also forms the basis of the techniques to be implemented in the radioactive source calibration of the Phase I upgrade of the Hadron Endcap Calorimeters during Year End Technical Stop 2017-2018. The test validates the automated radioactive source calibration concept for scintillator-based calorimeters and provides insight on the applicability of this concept to large-scale detector systems.